Regularization of Complex SAR Images Using Markov Random Fields

This paper presents despeckling and information extraction using non-quadratic regularization. The novelty of this paper is that instead of the Gaussian prior model a Gauss-Markov random field model is chosen, because it can efficiently model textures in the images. The iterative procedure consists of noise-free image and texture parameter. The experimental results show that the proposed method satisfactorily removes noise form synthetic and real SAR images and is comparable with the state of the art methods using objective measurements on synthetic SAR images.